Indoor position fingerprint positioning algorithm based on APSO-LSSVM

Abstract

The invention provides an APSO-LSSVM-based indoor position fingerprint positioning algorithm, and belongs to the technical field of indoor positioning, and the method comprises the following steps: 1, collecting data, and carrying out the preprocessing of the data; step 2, constructing an LSSVM (least square support vector machine) model; step 3, optimizing LSSVM model parameters by using an improved parameter optimization algorithm; 4, training and verifying the optimized model; and 5, performing indoor positioning on an indoor object by using the trained model. The improved parameter optimization algorithm APSO is used for optimizing parameters C and sigma in the positioning model, and the optimal parameter is determined. The LSSVM reduces the calculation complexity, shortens the training time, and is higher in calculation efficiency, so that in indoor positioning, the condition that the number of training samples is large can be better processed, and the requirements of an indoor positioning algorithm can be better met.

Description

Technical Field

[0001] The present invention relates to the field of indoor positioning technology, and in particular to an indoor position fingerprint positioning algorithm based on APSO-LSSVM. Background Art

[0002] With the popularization of smart devices and the development of mobile Internet, people's demand for indoor and outdoor positioning services is growing. Traditional GPS positioning cannot provide satisfactory accuracy in indoor environments due to signal obstruction and other reasons. Therefore, location fingerprint positioning technology based on Wi-Fi signal strength has attracted widespread attention due to its high accuracy and low cost. However, the existing location fingerprint positioning methods have problems such as high model training complexity and great environmental impact on positioning accuracy. Although the traditional LSSVM algorithm can provide good positioning accuracy, its parameter selection depends on experience or manual adjustment, which increases the difficulty of model training and may affect the final positioning effect. In addition, environmental factors such as people walking and changes in equipment layout may cause signal fluctuations, thereby reducing positioning accuracy. Therefore, it is necessary to design an indoor location fingerprint positioning algorithm based on APSO-LSSVM. Summary of the invention

[0003] The purpose of the present invention is to provide an indoor position fingerprint positioning algorithm based on APSO-LSSVM to solve the technical problems of insufficient indoor positioning accuracy and slow data processing.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] An indoor location fingerprint positioning algorithm based on APSO-LSSVM, the method comprises the following steps:

[0006] Step 1: Collect data and preprocess the data;

[0007] Step 2: Build LSSVM model;

[0008] Step 3: Use the improved parameter optimization algorithm to optimize the LSSVM model parameters;

[0009] Step 4: Train and verify the optimized model;

[0010] Step 5: Use the trained model to perform indoor positioning of indoor objects.

[0011] Furthermore, the process of collecting data in step 1 is as follows: obtaining the signal strength indicator RSSI from different access points APs according to Wi-Fi, marking the coordinates of each location and the signal strength received by each wireless access point at several known locations in the target area, determining one or several paths that pass through the entire target area to ensure that all set locations are covered, walking along the planned path, and starting the data collection tool at the same time. At each predefined location, the coordinates and corresponding signal strength of the location are recorded, and data at the same location are sampled several times at different times to avoid the impact of environmental changes. Each data should contain a clear latitude and longitude coordinate, signal strength and time.

[0012] Furthermore, the preprocessing process in step 1 is as follows: the standard deviation between each data point and the mean is calculated. If the standard score of a point is greater than the set threshold, it is considered an outlier, and a linear regression model is used to predict and fill in the missing values.

[0013] Furthermore, the specific process of step 2 is: setting the selection function, which is as follows:

[0014] K(x,x′)=exp(-γ||xx′|| 2 )

[0015] Where x and x′ are input feature vectors, and γ is a hyperparameter that controls the width of the selection function;

[0016] Set the search space, set the complexity σ, σ is the regularization parameter, and set the penalty parameter C.

[0017] Furthermore, the specific process of step 3 is as follows: define the fitness function, which is used to evaluate the performance of the LSSVM model under a given parameter combination, use cross-validation to evaluate the model, and select the mean square error as the fitness value, set the initialization particles of APSO, update the particle position and velocity, and update the global optimal solution. For each iteration, update the position and velocity of each particle, and update the personal best position and personal best fitness according to the fitness function, prepare the data set and run APSO to find the optimal parameters.

[0018] Furthermore, the specific process of step 4 is as follows: train the LSSVM model with the best parameters found, use k-fold cross validation to evaluate the generalization ability of the model to ensure that there is no overfitting, use the validation set to evaluate the performance, first make predictions on the validation set, calculate the performance indicators on the validation set, and use the test set to evaluate the performance.

[0019] Furthermore, the real-time RSSI value of the location of the object to be located is obtained in real time, including the received signal strength information of several wireless access points. The RSSI value collected in real time is preprocessed in the same way as the training data, and is standardized using the StandardScaler tool. The preprocessed real-time data is predicted using the trained LSSVM model, and the x-coordinate and y-coordinate or longitude and latitude information of the located object on the two-dimensional plane is output.

[0020] The present invention has the following beneficial effects due to the adoption of the above technical solution:

[0021] The improved parameter optimization algorithm APSO of the present invention optimizes the parameters C and σ in the positioning model to determine the optimal parameters. LSSVM reduces the computational complexity, shortens the training time, and has relatively high computational efficiency. Therefore, in indoor positioning, it can better handle the situation of a large number of training samples and can better meet the requirements of indoor positioning algorithms. Parameter C and kernel parameter σ are two important undetermined parameters in the LSSVM model based on the kernel function. They determine the learning performance of LSSVM. If the parameter selection is unreasonable, it will directly affect the positioning accuracy. Compared with other optimization algorithms, the particle swarm optimization algorithm requires fewer parameters to be adjusted, has obvious advantages, reduces the computational complexity, and minimizes the influence of noise such as complex indoor environments. However, since PSO is prone to fall into the problem of local optimality, it cannot guarantee global optimization. Therefore, an adaptive particle swarm optimization algorithm (APSO) is proposed to improve the efficiency of parameter optimization in model construction, avoid the subjectivity of parameter selection of the prediction model, and effectively improve the prediction accuracy of the LSSVM model. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flow chart of the method of the present invention;

[0023] Figure 2 It is a comparison diagram of positioning errors of the present invention;

[0024] Figure 3 It is the object positioning coordinate diagram of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and preferred embodiments. However, it should be noted that many details listed in the specification are only for the purpose of enabling the reader to have a thorough understanding of one or more aspects of the present invention, and these aspects of the present invention can be implemented even without these specific details.

[0026] like Figure 1 As shown, the indoor position fingerprint positioning algorithm based on APSO-LSSVM, the method includes the following steps:

[0027] Step 1: Collect data and pre-process the data. According to Wi-Fi, obtain the signal strength indicator RSSI from different access points APs. At several known locations in the target area, mark the coordinates of each location and the signal strength received by each wireless access point. Determine one or several paths that pass through the entire target area to ensure that all set locations are covered. Walk along the planned path and start the data collection tool at the same time. At each predefined location, record the coordinates and corresponding signal strength of the location. Sample the data of the same location several times at different times to avoid the impact of environmental changes. Each data should contain a clear latitude and longitude coordinate, signal strength and time. Calculate the standard deviation between each data point and the mean. If the standard score of a point is greater than the set threshold, it is considered an outlier. Use a linear regression model to predict and fill in the missing values. Select the most useful features for positioning from the collected data. Use statistical analysis or machine learning methods to select the most critical features for positioning from the original data set, remove redundant information, and simplify the calculation process.

[0028] Step 2: Build the LSSVM model. Set the selection function as follows:

[0029] K(x,x′)=exp(-γ||xx′|| 2 )

[0030] Where x and x′ are input feature vectors, and γ is a hyperparameter that controls the width of the selection function;

[0031] Set the search space, set the complexity σ, σ is the regularization parameter, and set the penalty parameter C.

[0032] Step 3: Optimize the parameters of the LSSVM model using the improved parameter optimization algorithm. Define the fitness function, which is used to evaluate the performance of the LSSVM model under a given parameter combination. Use cross-validation to evaluate the model and select the mean square error as the fitness value. Set the initial particles of APSO, update the particle position and velocity, and update the global optimal solution. For each iteration, update the position and velocity of each particle, and update the personal best position and personal best fitness according to the fitness function. Prepare the data set and run APSO to find the best parameters.

[0033] Step 4: Train and validate the optimized model. Use the best parameters found to train the LSSVM model, use k-fold cross validation to evaluate the generalization ability of the model to ensure that there is no overfitting, and use the validation set to evaluate the performance. First, make predictions on the validation set, calculate the performance indicators on the validation set, and use the test set to evaluate the performance. Initialize a set of random solutions to represent the parameter set of LSSVM; then, use the APSO algorithm to iteratively search for the optimal parameter combination, and adjust the inertia weight coefficient so that the algorithm can better balance the global search ability and local convergence speed; finally, apply the obtained optimal parameters to the LSSVM model to complete the model training.

[0034] Step 5: Use the trained model to locate indoor objects indoors. Obtain the real-time RSSI value of the location of the object to be located in real time, including the signal strength information of several wireless access points. Perform the same preprocessing as the training data on the RSSI value collected in real time, use the StandardScaler tool to perform standardization, use the trained LSSVM model to predict the preprocessed real-time data, and output the x-coordinate and y-coordinate or longitude and latitude information of the located object on the two-dimensional plane. Input the characteristic values ​​such as the Wi-Fi signal strength of the location to be located into the trained LSSVM model to obtain the prediction result, that is, the specific coordinates of the location.

[0035] This method uses the least squares support vector machine (LSSVM) to build a positioning model, and proposes an improved parameter optimization algorithm APSO to optimize the parameters C and σ in the positioning model and determine the optimal parameters. LSSVM reduces the computational complexity, shortens the training time, and has high computational efficiency. Therefore, in indoor positioning, it can better handle the situation of a large number of training samples and can better meet the needs of indoor positioning algorithms. Parameter C and kernel parameter σ are two important undetermined parameters in the LSSVM model based on the kernel function. They determine the learning performance of LSSVM. If the parameter selection is unreasonable, it will directly affect the positioning accuracy. Compared with other optimization algorithms, the particle swarm optimization algorithm requires fewer parameters to be adjusted, has obvious advantages, reduces the computational complexity, and minimizes the influence of noise such as complex indoor environments. However, since PSO is prone to fall into the problem of local optimality, it cannot guarantee global optimization. Therefore, an adaptive particle swarm optimization algorithm (APSO) is proposed to improve the efficiency of parameter optimization in model construction, avoid the subjectivity of parameter selection of the prediction model, and effectively improve the prediction accuracy of the LSSVM model.

[0036] like Figure 2 As shown, the error of the positioning algorithm of the present application is smaller than that of the existing KNN, WKNN, etc., and accurate indoor positioning can be achieved. The positioning error is smaller than that of other existing technologies and is more accurate.

[0037] like Figure 3 It is the specific positioning of the position of the object in the room and the estimated position. By comparing the real position coordinates with the estimated position coordinates, it can be seen that the error is at the mm level, achieving high-precision positioning.

[0038] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. Indoor location fingerprint positioning algorithm based on APSO-LSSVM, characterized by: The method comprises the following steps: Step 1: Collect data and preprocess the data; Step 2: Build LSSVM model; Step 3: Use the improved parameter optimization algorithm to optimize the LSSVM model parameters; Step 4: Train and verify the optimized model; Step 5: Use the trained model to perform indoor positioning of indoor objects.

2. According to the APSO-LSSVM-based indoor location fingerprint positioning algorithm of claim 1, it is characterized by: The process of collecting data in step 1 is as follows: obtain the signal strength indicator RSSI from different access points APs according to Wi-Fi, mark the coordinates of each location and the signal strength received by each wireless access point at several known locations in the target area, determine one or several paths that pass through the entire target area to ensure that all set locations are covered, walk along the planned path, and start the data collection tool at the same time. Every time you arrive at a predefined location, record the coordinates of the location and the corresponding signal strength. Sample the data of the same location several times at different times to avoid the impact of environmental changes. Each data should contain a clear latitude and longitude coordinate, signal strength and time.

3. The indoor position fingerprint positioning algorithm based on APSO-LSSVM according to claim 1 is characterized in that: The preprocessing process in step 1 is: calculate the standard deviation between each data point and the mean, if the standard score of a point is greater than the set threshold, it is considered an outlier, and a linear regression model is used to predict and fill in the missing values.

4. The indoor position fingerprint positioning algorithm based on APSO-LSSVM according to claim 1 is characterized in that: The specific process of step 2 is: set the selection function, as shown in the following formula: K(x,x′)=exp(-γ||x-x′|| 2 ) Where x and x′ are input feature vectors, and γ is a hyperparameter that controls the width of the selection function; Set the search space, set the complexity σ, σ is the regularization parameter, and set the penalty parameter C.

5. The indoor position fingerprint positioning algorithm based on APSO-LSSVM according to claim 1 is characterized in that: The specific process of step 3 is as follows: define the fitness function, which is used to evaluate the performance of the LSSVM model under a given parameter combination, use cross-validation to evaluate the model, and select the mean square error as the fitness value, set the initialization particles of APSO, update the particle position and velocity, and update the global optimal solution. For each iteration, update the position and velocity of each particle, and update the personal best position and personal best fitness according to the fitness function, prepare the data set and run APSO to find the optimal parameters.

6. The indoor position fingerprint positioning algorithm based on APSO-LSSVM according to claim 1 is characterized in that: The specific process of step 4 is as follows: train the LSSVM model with the best parameters found, use k-fold cross validation to evaluate the generalization ability of the model to ensure that there is no overfitting, use the validation set to evaluate the performance, first make predictions on the validation set, calculate the performance indicators on the validation set, and use the test set to evaluate the performance.

7. The indoor position fingerprint positioning algorithm based on APSO-LSSVM according to claim 1 is characterized in that: The real-time RSSI value of the location of the object to be located is obtained in real time, including the received signal strength information of several wireless access points. The real-time collected RSSI value is preprocessed in the same way as the training data, and the StandardScaler tool is used for standardization. The preprocessed real-time data is predicted using the trained LSSVM model, and the x-coordinate and y-coordinate or longitude and latitude information of the located object on the two-dimensional plane is output.

Images